INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

been

-0.81

 palate

-0.66

 Compos

-0.62

oda

-0.61

akin

-0.60

ometown

-0.60

Joined

-0.60

eling

-0.60

ceived

-0.59

finished

-0.59

POSITIVE LOGITS

 briefs

0.71

 ARTICLE

0.67

Ups

0.66

atre

0.64

BILITIES

0.62

unin

0.62

 Oliv

0.62

otine

0.62

 Baxter

0.61

obin

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.